7 Efficient solutions to autonomous mapping and navigation problems版本

《7 Efficient solutions to autonomous mapping and navigation problems版本》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、EfficientSolutionstoAutonomousMappingandNavigationProblemsStefanBernardWilliamsAthesissubmittedinfulfillmentoftherequirementsforthedegreeofDoctorofPhilosophyAustralianCentreforFieldRoboticsDepartmentofMechanicalandMechatronicEngineeringTheUniversityofSydneySeptember

2、2001DeclarationThisthesisissubmittedtoTheUniversityofSydneyinfulfillmentoftherequirementsforthedegreeofDoctorofPhilosophy.Thisthesisisentirelymyownworkand,exceptwhereotherwisestated,describesmyownresearch.StefanBernardWilliamsAustralianCentreforFieldRoboticsTheUni

3、versityofSydneyCopyright
c2001StefanBWilliamsAllrightreservediiiAbstractStefanBernardWilliamsDoctorofPhilosophyTheUniversityofSydneySeptember2001EfficientSolutionstoAutonomousMappingandNavigationProblemsThisthesisdealswiththeSimultaneousLocalisationandMappingalgori

4、thmasitpertainstothedeploymentofmobilesystemsinunknownenvironments.SimultaneousLocalisationandMapping(SLAM)asdefinedinthisthesisistheprocessofconcurrentlybuildingupamapoftheenvironmentandusingthismaptoobtainimprovedestimatesofthelocationofthevehicle.Inessence,thev

5、ehiclereliesonitsabilitytoextractusefulnavigationinfor-mationfromthedatareturnedbyitssensors.Thevehicletypicallystartsatanunknownlocationwithnoaprioriknowledgeoflandmarklocations.Fromrelativeobservationsoflandmarks,itsimultaneouslycomputesanestimateofvehiclelocat

6、ionandanestimateoflandmarklocations.Whilecontinuinginmotion,thevehiclebuildsacompletemapoflandmarksandusesthesetoprovidecontinuousestimatesofthevehiclelocation.Thepo-tentialforthistypeofnavigationsystemforautonomoussystemsoperatinginunknownenvironmentsisenormous.

7、OnesignificantobstacleontheroadtotheimplementationanddeploymentoflargescaleSLAMalgorithmsisthecomputationaleffortrequiredtomaintainthecorrelationinforma-tionbetweenfeaturesinthemapandbetweenthefeaturesandthevehicle.PerformingtheupdateofthecovariancematrixisofO(n3)f

8、orastraightforwardimplementationoftheKalmanFilter.InthecaseoftheSLAMalgorithm,thiscomplexitycanbereducedtoO(n2)giventhesparsenatureoftypicalobservations.Evenso